Geneva has the data. It's missing the AI plumbing.

Geneva packs a unique density of organizations processing massive document volumes. The UN, WHO, and WTO produce reports in six languages. Private banks on Rue du Rhône handle thousands of compliance pages every quarter. Commodity trading firms analyze contracts, market reports, and shipping documents in three or four languages. All of this content could be processed by AI. Almost none of it is.

The main blocker isn't technical. It's confidentiality. Since the nFADP came into effect in September 2023, Swiss companies have one more reason to be cautious with public AI tools. Sending client data to ChatGPT is a legal risk. Banks know it, fiduciaries know it, law firms know it. The result: nothing happens. Or a POC gets built that carefully avoids touching real data, proving nothing.

In finance, the needs are concrete. Automating KYC/AML screening across thousands of transactions. Generating risk reports from structured and unstructured data. Extracting key clauses from 80-page trading contracts. Analysts do this manually, or with 2010-era tools that don't understand context.

Geneva has CERN at its doorstep and EPFL an hour away by train. The AI research ecosystem is right there. But between a transformer paper and a regulation-compliant AI deployment inside a private bank, there's a canyon. The Big Four sell seven-figure digital transformation programs. Geneva SMEs and family offices watch from a distance, convinced AI requires enormous budgets. It doesn't, as long as you target the right use case.

What I do for Geneva businesses

AI audit: find the high-ROI cases in your context

You run a fiduciary and spend hours extracting numbers from PDF documents for tax filings? You're in commodity trading and your analysts reread 100-page contracts to locate three clauses? I look at your processes, your data, your confidentiality constraints. I identify the cases where AI delivers a measurable gain. And when a Python script or an n8n automation is enough, I say so. Not every problem needs a language model.

Private LLM integration: AI without exposing your data

The critical point in Geneva is confidentiality. I deploy AI solutions that run on private APIs or models hosted in Europe. A RAG system connected to your internal document base so your team finds information in 10 seconds instead of digging through SharePoint for 20 minutes. An assistant that pre-analyzes your compliance reports and flags anomalies. Multilingual FR/EN/DE workflows that process your documents without a single data point leaving your perimeter. No self-service ChatGPT. Confined, auditable AI tools compliant with the nFADP.

AI training: get your teams operational

Your financial analysts, lawyers, and compliance officers don't need a deep learning course. They need to know how to use AI tools effectively and safely. How to detect when the model hallucinates. Which data can be submitted, which is off-limits. How to write a prompt that produces a usable result. I train on your real use cases, with your documents, in your environment. No generic slides about the history of artificial intelligence.

How it works

1. Initial call (free, 30 min)

We talk about your situation. Your tools, your pain points, your regulatory constraints. No slides, no sales pitch. If AI isn't the right answer for your case, I tell you. Nobody's time gets wasted.

2. AI maturity audit

I map your processes and data flows. I identify high-ROI use cases that are realistic with your current stack and confidentiality requirements. You get a clear deliverable: what to do, in what order, with an estimated budget and compliance guarantees.

3. POC then deployment

We start small. A working prototype on a concrete use case, tested by your teams with real data (not a dummy dataset). If it works and compliance is validated, we move to deployment. No six-month development tunnel before seeing a result.

4. Training and autonomy

I train the end users. Concrete stuff: how to use the tool, how to write good prompts, how to spot when the AI is wrong, which data to never submit. When I leave, your team runs it.

Tech stack

OpenAI APIClaude APILangChainRAGPythonn8nGoogle Apps Script

Frequently asked questions

You're based in Tbilisi. How does remote work for Geneva companies?

The audit happens over video calls, code is delivered continuously via Git, and I'm on the European timezone (Georgia is UTC+4, two hours ahead of Switzerland). For Geneva organizations used to working with distributed teams (which is the norm in international orgs), it's seamless. If an on-site workshop is needed, I travel. But most integration work is code, and code doesn't care where the keyboard is.

What about data confidentiality and Swiss law?

It's the first thing we address. I never route your data through public AI tools without your explicit approval. The solutions I deploy use private APIs with European processing guarantees, or models hosted within your infrastructure. I work with your legal team to validate nFADP compliance before any deployment. If your case requires Swiss-hosted infrastructure, we find the right architecture.

How much does an AI project cost?

Depends on the scope. An AI maturity audit for an SME or family office is a few days of work. A full LLM integration with a RAG system and team training takes a few weeks. The initial 30-minute call is free and exists to scope the need. I bill by day or by project, no surprises. And if the ROI isn't there, I tell you before we start.

We're a small firm (10-50 people). Is AI relevant for us?

Often more so than for a large corporation, proportionally. A 15-person fiduciary that automates data extraction from client documents saves hours every week. A 20-person consultancy that deploys a RAG system on its internal knowledge base speeds up every consultant's work. AI isn't reserved for SMI-listed companies. The key is targeting one specific use case with measurable ROI, not trying to do everything at once.